Assess list of plates in CPJUMP1 with known issues to see if assuming they affect the results alters any conclusions

[bethac07]

https://github.com/jump-cellpainting/2021_Chandrasekaran_submitted/issues/46#issuecomment-1103063970

[bethac07]

8 experiment notebooks use CPJUMP1 plates:

Three assess the same plates repeatedly, so should not be affected by this

• AssessBleaching
• AssessFOVs
• AssessTimedelay

Five look across plates, and thus might be affected

• AssessCompoundsInCas9
• AssessDensity
• AssessDrugSelection
• AssessTarget1vsTarget2
• AssessTreatmentTimepoint

[bethac07]

AssessCompoundsInCas9

Plates included

``` experiment_dict = { 'A549':{ 'Compounds':{ 24:['BR00116991','BR00116992','BR00116993','BR00116994'], 48:['BR00117015','BR00117016','BR00117017','BR00117019']}, 'ORF':{ 48:['BR00117020','BR00117021'], 96:['BR00118050','BR00117006']} }, 'A549-Cas9':{ 'Compounds':{ 48:['BR00117050','BR00117051','BR00117052','BR00117053']}, 'CRISPR':{ 96:['BR00118041','BR00118042','BR00118043','BR00118044'], 144:['BR00117000','BR00117003','BR00117004','BR00117005']} } } ```

Potential issues in these plates -

• Increase in WGA - 6991-6994 (All A549-Compounds-24)
• Possible focal issues - 6993 (one in A549-Compounds-24)
• Increase in Phalloidin - 7006, 8041-8044, 8050 (All in A549-ORF-96, All in A549-CRISPR-96)

A priori expectations if these problems are critical -

• A549-Compounds-24 may have a slightly lower percent replicating than otherwise expected, due to the focal issues
• A549-Compounds-24 may not match well to either 48 hour compound plate very well, but we really only care about the DIFFERENCE between the two 48 hour compound plates
• 48 hour compound & ORF matching might be poorer to 96 hour CRISPR or ORF matching than it is to each other or to 144 hour CRISPR

Figures:

Conclusions -

• A549-Compounds-24 has a slightly lower but still REALLY high PR (Fig 1)
• A549-Compounds-24 matching to the two Compounds-48 plates is comparable to the matching of the two 48 hour plates to each other (Fig 2, top left, red vs purple dots), so no evidence of issues
• All cross modality matches are really low, but no evidence that 24 hour containing matches (red, blue, green) or 96 hour-to-another-timepoint matches (blue, orange, pink) are worse than other matches (purple, yellow, brown)

No demonstrable effects

[bethac07]

AssessTreatmentTimepoint

Plates included

``` 'A549':{ 'Compounds':{ 24:['BR00116991','BR00116992','BR00116993','BR00116994'], 48:['BR00117015','BR00117016','BR00117017','BR00117019']}, 'ORF':{ 48:['BR00117020','BR00117021'], 96:['BR00118050','BR00117006']}, 'CRISPR':{ 96:['BR00118041','BR00118042','BR00118043','BR00118044'], 144:['BR00117000','BR00117003','BR00117004','BR00117005']} }, 'U2OS':{ 'Compounds':{ 24:['BR00116995','BR00117024','BR00117025','BR00117026'], 48:['BR00117010','BR00117011','BR00117012','BR00117013']}, 'ORF':{ 48:['BR00117022','BR00117023'], 96:['BR00118039','BR00118040']}, 'CRISPR':{ 96:['BR00118045','BR00118046','BR00118047','BR00118048'], 144:['BR00116996','BR00116997','BR00116998','BR00116999' ]} } } ```

Potential issues in these plates -

• Increase in WGA - 6991-6999, 7024-7026 (All Compounds-24, U2OS-CRISPR-144)
• Possible focal issues - 6993 (one in A549-Compounds-24)
• Possible cell loss issues -6995 (one in U2OS-Compounds-24)
• Possible drift issues- 6996 (one in U2OS-CRISPR-144)
• Possible extra mitotracker incubation in a subset of wells - 7010 OR 7013 (one in U2OS-Compounds-48, but not certain which)
• Increase in Phalloidin - 7006, 8039-8048, 8050 (All in ORF-96, All in CRISPR-96)

A priori expectations if these problems are critical -

• Possible lower Percent Replicating in Compounds-24, U2OS-Compounds-48, and/or U2OS-CRISPR-144
• Possible lower Percent Matching for Compounds-24 or U2OS-CRISPR-144 to batches other than these
• Possible lower Percent Matching for ORF-96 or CRISPR-96 to batches other than these

Figures:

Conclusions -

• No reason to suspect lower percent replicatings for these batches (Figure 1)
• 24 vs 48 hour matching to ORFs or CRISPRs (top middle and top right panels, compare red to purple, blue to orange, and blue to yellow for A549 only) is worse, but hard to say if it's a timepoint (previous data suggests 24 hrs is worse than 48 hours or a staining issue. Green to yellow (24 vs 48 hour compounds in 144) in U2OS is worse, not better than it is for A549. Matching of U2OS-CRISPR-144 to ORFs (bottom middle panel) relative to 96 hrs (compare orange to yellow and brown to pink) is worse than in A549.
• No evidence that 48 hr compounds match worse to 96 hour ORFs than 48 hour ORFs (top center, compare purple to orange), possible but weak evidence for 48 hour matching to 96 hr CRISPRs in A549 (can't do U2OS due to WGA issue; top right, compare orange to yellow), possible but weak evidence for 96-96 hour matching in the ORF-to-CRISPR for A549 being better than 96-to-other matching (bottom middle, compare brown to orange and pink)

Possible effects, though I'm not convinced; could rerun these profiles dropping the AGP channel (covers both WGA and Phalloidin) to see if the conclusions change. - Edited to add - now running in #9

[bethac07]

AssessDensity

Plates included

``` experiment_dict = { 'A549':{ 'Compounds':{ 24:['BR00116991','BR00116992','BR00116993','BR00116994'], 48:['BR00117015','BR00117016','BR00117017','BR00117019']}, 'ORF':{ 48:['BR00117020','BR00117021'], 96:['BR00118050','BR00117006']}, 'CRISPR':{ 96:['BR00118041','BR00118042','BR00118043','BR00118044'], 144:['BR00117000','BR00117003','BR00117004','BR00117005']} }, 'A549 80%':{ 'Compounds':{ 48:['BR00117008','BR00117009']} }, 'A549 120%':{ 'Compounds':{ 48:['BR00117054','BR00117055']} } } ```

Potential issues in these plates -

• Increase in WGA - 6991-6994, (Compounds-24)
• Possible focal issues - 6993 (one in A549-Compounds-24)
• Increase in Phalloidin - 7006, 8041-8044, 8050 (All in ORF-96, All in CRISPR-96)

A priori expectations if these problems are critical -

• Possible lower Percent Replicating in Compounds-24
• Possible lower Percent Matching for Compounds-24 to other batches
• Possible lower Percent Matching for ORF-96 or CRISPR-96 to batches other than these

Conclusions: None of the 48 hr compound plates are impacted, and we only care about the differences between performances of these batches, so this experiment should not be affected

[bethac07]

AssessDrugSelection

Plates included

``` experiment_dict = { 'A549':{ 'Compounds':{ 24:['BR00116991','BR00116992','BR00116993','BR00116994'], 48:['BR00117015','BR00117016','BR00117017','BR00117019']}, 'ORF':{ 48:['BR00117020','BR00117021'], 96:['BR00118050','BR00117006']}, 'CRISPR':{ 96:['BR00118041','BR00118042','BR00118043','BR00118044'], 144:['BR00117000','BR00117003','BR00117004','BR00117005']} }, 'A549+Selection':{ 'ORF':{ 96:['BR00118049']}, 'CRISPR':{ 144:['BR00117001','BR00117002']} } } ```

Potential issues in these plates -

• Increase in WGA - 6991-6994 (Compounds-24)
• Possible focal issues - 6993 (one in Compounds-24)
• Increase in Phalloidin - 7006, 8041-8044, 8049-8050 (All in ORF-96, All in CRISPR-96)

A priori expectations if these problems are critical -

• Possible lower Percent Replicating in Compounds-24
• Possible lower Percent Matching for Compounds-24 to other batches
• Possible lower Percent Matching for ORF-96 or CRISPR-96 to batches other than these

Conclusions - We only care about the difference between A549 +/- selection here, and all batches behave the same across that variable, so this experiment should not be affected

[bethac07]

AssessTarget1VsTarget2

Plates included

``` experiment_dict = { 'Target1':['BR00117010','BR00117011','BR00117012','BR00117013'], 'Target2':['BR00121434','BR00121435','BR00121440'] } ```

Potential issues in these plates -

• Possible extra mitotracker incubation in a subset of wells - 7010 OR 7013 (one in Target1, but not certain which)

A priori expectations if these problems are critical -

• Possible lower Percent Replicating and/or cross plate Percent Replicating for Target1

A priori expectations if these problems are critical -

• Possible lower Percent Replicating in Compounds-24, U2OS-Compounds-48, and/or U2OS-CRISPR-144
• Possible lower Percent Matching for Compounds-24 or U2OS-CRISPR-144 to batches other than these
• Possible lower Percent Matching for ORF-96 or CRISPR-96 to batches other than these

Figures:

Conclusions - Impossible to know for sure without rerunning to know which are the lower dots in cross-plate Percent Replicating, but since overall Replicating looks totally equal and cross-plate is close within each batch, no reason to doubt the conclusion that the plates are functionally equivalent.

[bethac07]

Fresh data, hot off the presses!

Top is our original figure, bottom is without AGP. Insomuch as we can conclude anything about matching in this experiment (we're doing a two-tailed match here, so "null" should be 10%) , I don't think overall our conclusions would change much with or without AGP. A couple of the relative positions are changed, but most are very small changes and split about equally between "move direction we'd expect if staining mistake was hurting matching" and "move opposite of what we'd expect". In any case, there's nothing here that contradicts any decisions.